基于三线性张量生成图像的方法  

Image Rendering Metheod Based on Tri-linear Tensor

作　　者：李保洲[1] 邓雁萍[1] 李介谷[1] 

机构地区：[1]上海交通大学图像处理与模式识别研究所,上海200030

出　　处：《数据采集与处理》2000年第3期320-323,共4页Journal of Data Acquisition and Processing

摘　　要：介绍了一种新的基于代数约束的图像生成方法 ,用于从两幅不同角度的参考图像生成另一幅新的视角图像 ,即二次投影。该方法在极线几何的基础上 ,通过研究物体在三个不同透视投影平面上对应点之间的相互关系 ,得到描述不同图像上对应点关系的一个三线性张量。由此张量和参考图像便能得到新图像和参考图像之间的关系 ,进而生成新的视角图像。和其他基于图像的二次投影方法相比 ,这里仅需要两幅参考图像 ,且实现方法简单快速。实验给出了令人较满意的效果。A new technique for the image rendering based on algebr aic constraint is proposed. It is used to render another novel view from two ref erence images of different locations,which is called re-projection. By studyin g the geometrical relation among the corresponding points on three different projective planes based on epipolar geometry, the tri-linear tensor constrainin g the corresponding points of different views is obtained. From this tensor and the corresponding relation between the reference images,the relation betwee n the novel view and the reference images can be obtained and the novel view thu s achieved. Comparing with other image re-projection methods,this method ne eds only two reference images and the implementation is easy and quick. Satisfac tory results are given in the experiment.

关 键 词：代数约束 图像生成 三线性张量 计算机图形学 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]